VIDEO
HIGHLIGHT
Log InSign up
Rewriting the MT Playbook with Marco Trombetti
SummaryTranscriptChat

Rewriting the MT Playbook with Marco Trombetti

Introduction to AI Agents and Localization

  • Eddie Arta introduces Marco Tereti, CEO of Translated, discussing AI agents and their impact on the localization industry.
  • Marco expresses gratitude for being invited back and mentions the importance of industry events.
  • They discuss a recent release from Translated, highlighting its significance in the localization field.

Lara: Next Generation Machine Translation Technology

  • Marco explains that Lara combines large language models with translation models for improved fluency and accuracy.
  • The technology aims to reduce hallucination while maintaining flexibility in translations.
  • Lara was first introduced on November 4th, with updates following in February and April.

Expansion of Language Support

  • Translated has expanded support from 10 to 32 languages, enhancing capabilities significantly.
  • The new model allows real-time adaptation without retraining when errors are found during production.
  • Contextual understanding is improved, enabling better management of conversations between users.

Enhanced Features of Lara

  • Users can provide specific instructions for tasks like SEO directly within the translation process.
  • This integration reduces the need for multiple processing steps by centralizing functions within one model.
  • Results are generated quickly (in about 500 milliseconds), improving efficiency over traditional methods.

Team Plan: API and AI Agent Discussion

  • Marco introduces Team Plan as an API and AI agent, addressing confusion around these concepts in the industry.
  • Eddie acknowledges his own struggles with understanding AI agents at a high level.

Team Functionality in Localization

  • New team functionality allows centralized management of user accounts and terminology for enterprises.
  • Ensures all employees use consistent terminology, enhancing translation quality and safety.
  • Localization teams regain control over machine translation processes previously managed by tech departments.

Improving Internal Communication

  • Localization departments can share data with all employees, promoting their work within the organization.
  • Every professional translation enhances the internal model used for communication and understanding.
  • Employees gain access to powerful translation tools, improving overall experience and visibility.

API Enhancements

  • Introduction of a high-performance machine translation API designed for easy integration into workflows.
  • Optimized for processing large amounts of user-generated content at competitive costs with low latency.
  • Achieves average latencies around 500 milliseconds, suitable for real-time applications.

AI Agents Framework

  • AI agents are defined as AIs that interact with each other to perform complex tasks beyond simple transactions.
  • Recent developments have enabled protocols allowing AI agents to communicate effectively.
  • The release of an open-source protocol (MCP) has sparked interest in building more interactive AI tools.

Building Tools with MCP Protocol

  • MCP protocol enables developers to create tools that enhance AI capabilities through contextual information sharing.
  • Initial focus was on adding context for better translations; however, developers shifted towards creating utility tools instead.

The Rise of AI Agents and Protocols

Introduction to AI Agents

  • The speaker discusses executing codes for applications, leading to live projects that users engage with unexpectedly.
  • OpenAI, Microsoft, and Google announce support for the NCP protocol, indicating a shift towards unified communication among AI agents.

Functionality of AI Agents

  • Users can request tasks from AI agents, such as booking vacations without prior knowledge of details.
  • Multiple agents can collaborate (e.g., LLM for queries and booking agents like Expedia or Sky Scanner).

Current Limitations

  • There is no app store for these agents yet; visibility and availability are limited.
  • The MCP protocol has design flaws due to its origins in another application.

Experimentation with Agents

  • The speaker's team created an MCP agent named Lara to automate translation project management.
  • Lara simplifies localization by translating spreadsheets automatically based on user input.

Advantages of Using Lara

  • Lara provides automation in project management, reducing manual copy-pasting efforts.
  • Teams gain visibility through a team plan that enhances global communication within organizations.

Challenges in Agent Development

  • Limited number of high-quality agents exists despite many websites offering services.

Challenges in Digital Service Functionality

  • Need for more agents to perform various digital tasks effectively, reflecting real-life complexities.
  • Agents should communicate symmetrically; current limitations prevent full utilization of platforms like Entropic and Lara.
  • Current internet phase is competitive, hindering collaboration and optimization among services.

Concerns About Market Dynamics

  • Desire for open standards to allow broader access and usage of platforms like Entropic.
  • Financial focus is crucial to avoid market dilution; unpredictability remains a challenge in the industry.

Future Engagement and Community Feedback

  • Appreciation expressed for community support and updates on trends in localization.
  • Upcoming release of Lara encourages user feedback; aim to engage with the localization industry actively.

Conclusion of Discussion

  • Summary of insights shared about AI agents, protocols, and the technological landscape's evolution.
Channel: MultiLingual
Video description

Marco Trombetti, CEO and Co-Founder of Translated, discusses the latest updates to LARA—their large language model-powered translation system—and how it's reshaping expectations around adaptation, speed, and context in machine translation. Marco also walks us through the current state of AI agents, their interoperability challenges, and what MCP protocol adoption by major players means for the future. We explore how localization teams can leverage LARA’s new Team Plan and API to enhance both internal workflows and enterprise-wide communication, and how the rise of agents could automate not just translation, but localization project management itself.